A Smart LSTM for Industrial Part Conformity: SME Material-Data Based Decision-Making

In predictive analysis for industrial parts, numerous studies have demonstrated the effectiveness of utilizing raw material data for quality prediction in monitoring systems. By analyzing and defining the composition, properties, and characteristics of raw materials, predictive models can optimize system performance and decision-making processes. In the context of small and medium enterprises (SMEs) like SCODER, limited resources constrain data collection for decision-making due to factors such as limited data availability, smaller production volumes, and less advanced traceability, leading to variability in data quality and structure. This research explores the impact of material characteristics on industrial part conformity with scarce and sparse data. Material impurities significantly affect the quality and stability of industrial machined parts. Some machine learning and AI techniques address the complexity of analyzing material datasets to identify patterns that lead to deficiencies in machined parts. We propose a smart long short-term memory (LSTM) model for predicting part conformity in industrial machined parts. This work focuses on analyzing raw material data to study machining stability and understand the dependencies between material data and part quality holistically. Our approach serves as a tool for ensuring quality production and industrial processing stability, leveraging data collected from machining parts at SCODER, a French SME specializing in ultraprecise stamping for automotive applications.